A. Mycobacteria and Disease
Mycobacterium tuberculosis (MTB) is the causative agent of the most common infectious disease in the world today, tuberculosis (TB). The World Health Organization (WHO) reported that 1.7 billion people (or approximately one-third of the world's population) are currently, or have been, infected by tuberculosis (Kochi, A. Tubercle 72:1-6 (1991)). The incidence of MTB infection is occurring at an increasing rate with 8 million new cases worldwide in 1991 (Sudre, P. et al., Bull. W.H.O. 70:149-159 (1992)), and the WHO estimates that 88.2 million people will contract TB during the 1990's and approximately 3 million people will die annually during this time period (Morbidity and Mortality Weekly Report 42 (No. 49):961-964 (1993)). In the United States the Centers for Disease Control and Prevention (CDC) recorded 26,673 cases in 1992 (Morbidity and Mortality Weekly Report 42:696-703 (1993)), and it is estimated that 10 to 15 million people in the U.S. have latent infections (Morbidity and Mortality Weekly Report 39(RR-8):9-12 (1990)).
The importance of Mycobacteria of the MAC complex, (primarily M. avium and M. intracellulare), as human pathogens was recently reviewed by Inderlied, C. B. et al., Clin. Microbiol. Rev. 6:266-310 (1993). MAC complex infections have been on the rise owing to their occurrence as opportunistic pathogens in AIDS patients. Approximately 43% of AIDS patients, with advanced stages of the disease, present with disseminated MAC infections (Nightingale et al., Jour. Infect. Dis. 165:1082-1085 (1992)). The WHO estimates that today approximately 3 million people have developed AIDS, approximately 15 million have been infected with the human immunodeficieney virus (HIV), and by the year 2000 the number infected could climb to approximately 40 million (World Health Organization (document WHO/GPA/CNP/EVA/93.1) Global Programme on AIDS (1993)). In addition to AIDS related infections, M. paratuberculosis, a subspecies of M. avium (Thorel, M. F. et al., Int. J. Syst. Bacteriol. 40:254-260 (1990)), is thought to be associated with Crohn's disease, an inflammatory disease of the bowel (Chiodini, R. J. Clin. Micro. Rev. 2:90-117 (1989)).
Mycobacterial infections are also a problem in animals. M. paratuberculosis also causes bowel inflammations in ruminants (Thoen, C. O. et al., Rev. Infect. Dis. 3:960-972 (1981)). This is more commonly known as Johne's disease. Cattle that test positive for Johne's are culled and destroyed. In the state of Wisconsin, where approximately one-third of the herds are infected (Collins, M. T., Hoard's Dairyman Feb 10:113 (1991)), the financial impact was estimated at $52 million in 1983 (Arnoldi, J. M. et al., Proceedings, 3rd Int. Symp. World Assoc. Vet. Lab. Diag. 2:493-494 (1983)). The incidence amongst herds nationwide typically ranges between 3% and 18% (Merkal, R. S. et al., J. Am. Vet. Med. Assoc. 190:676-680 (1987)). The financial impact of this one disease on the dairy industry exceeds $1.5 billion annually (Whitlock, R. Proceedings of the Third International Colloquium on Paratuberculosis, pp.514-522 (1991); Whitlock, R. et al., Proceedings of the 89th Annual Meeting of the United States Animal Health Association, pp.484-490 (1985)).
In addition to the organisms discussed above, a wide variety of Mycobacteria are also considered human pathogens, including Mycobacterium leprae, Mycobacterium kansasii, Mycobacterium marinum, Mycobacterium fortuitum complex, and many others. Wayne, L. G. et al., Clin. Micro. Rev. 5:1-25 (1992) review the diversity of infections associated with this genus of microorganism. However, the magnitude and impact of these infections is not on the same scale as MTB complex and MAC infections. For example, leprosy is probably the most common within this category: them are an estimated 5.5 million cases of Mycobacterium leprae worldwide (Nordeen, S. K. et al., Int. J. Lepr. 63:282-287 (1993)). Taken as a whole, this group of organisms exacts a tremendous social cost.
B. Culture and Detection of Mycobacteria
The contemporary protocol(s) for the laboratory diagnosis of Mycobacterial infections are relatively slow. Extended incubations are required owing to the innate slow growth rate of these bacteria. Owing to this lengthy time to diagnosis, individuals suspected of infection are quarantined, or else pose significant risk to society in general.
In addition, laboratory confirmation of the diagnosis of Mycobacterial infections requires several cultures per patient sample. Each sample must be incubated up to eight weeks (sixteen weeks for M. paratuberculosis) before the sample can be reported negative. The need for multiple cultures of each suspected sample is due in part to the intermittent shedding of detectable numbers of Mycobacteria, and the loss of infectious organisms due to the harsh chemical decontamination used to inactivate saprophytic microorganisms. These procedures are inefficient and often kill the Mycobacteria they are attempting to extract. For example, processing by the recommended N-acetyl-L-cysteine-NaOH (NALC/NaOH) procedure (Kent, P. T. et al., "Public Health Mycobacteriology," in A Guide for the Level III Laboratory, U.S. Department of Health and Human Service, Centers for Disease Control, (1985) pp. 31-46) is known to kill 28%-33% of the existing Mycobacteria (Krasnow, I. et al., Am. J. Clin. Path. 45:352-355 (1966); Kubica, G. P. W. et al., Am. Rev. Resir. Dis. 87:775-779 (1963)). The advent of contemporary probe assays (Gonzalez, R. et al., Diag. Microbiol. Infect. Dis. 8:69-78 (1987)) that complement culture techniques has improved the time to diagnosis; however, there still exists room for significant improvement.
The combination of social importance and reliance on culture methods reveals a critical need for a Mycobacterial testing protocol that reduces turn around time and increases sensitivity. The isothermal scheme being commercialized by Gen-Probe, Inc. (San Diego, Calif.: Jonas, V. et al., J. Clin. Micro. 31:2410-2416 (1993)) and the polymerase chain reaction (PCR) both have the potential for single molecule detection (Higuchi et al., Nature (London) 332:543-546 (1988)). Furthermore, amplification and detection can be performed in approximately eight hours and the reagents do not add a significant cost. If available, an amplification assay could greatly enhance the speed and sensitivity of detection, and reduce the cost of Mycobacterial diagnosis (De Cresce, R. P. et al., Med. Laboratory Obs. 25:28-33 (1993)). The rapidity with which these technologies could potentially diagnose Mycobacterial infections would have a tremendous financial impact on society.
However, as described herein, researchers have encountered a plethora of problems in an effort to adapt these technologies, such as PCR amplification, to the detection of Mycobacteria. Especially it has not been possible to develop a protocol for the preparation of a sample for analysis in a manner that will (a) ensure amplification assay detection of a true positive result and also (b) not give false negative results. The variability encountered by researchers is exemplified by the study of Noordhoek, G. T. et al., J. Clin. Micro. 32:277-284 (1994). These authors describe a blind study in which seven laboratories participated. All laboratories used the same amplification system, but different processing and detection methodologies. The original summary of these results (Noordhoek, G. T. et al., N. Eng. J. Med. 329:2036 (1993) concisely shows that at low copy numbers (1000 copies), the correlation varied from 2% to 90%, with the average being 54%. As a result of these problems, there is still no available FDA approved TB-amplification kit.
C. Methods of Processing Mycobacterium Samples
A review of the scientific literature on Mycobacteria-nucleic acids amplification references relating to system design, sample processing techniques and clinical studies reveals the highly variable results that have precluded FDA approval of a TB-amplification kit.
Two different amplification schemes have been used; there are numerous PCR system designs and many clinical studies that focus on Mycobacterial infections, the vast majority on MTB. Typically, samples are processed for culture first and then subjected to amplification. Therefore, sample preparation for amplification can be viewed, in most cases, as an extension of the culture processing protocol.
There are several reasons the field has evolved in this manner. First, obtaining clinical specimens is difficult. Individuals diagnosed with MTB are invariably started on drug therapy upon diagnosis. Second, processing MTB specimens requires specialized containment facilities and appropriately trained technicians. Third, it is the easiest way to obtain "culture correlation" results; that is, a correlation of amplification positive and negative results with those that were positive or negative in culture. Consequently, researchers have typically processed the sample for culture and then used protocols that "further" process the sediment for amplification. In this way, actual clinical specimens can be used, containment is not breached, work flow is not interrupted, patient care is not compromised and correlation to contemporary protocols is feasible. This "further" processing referred to above has involved a wide variety of sample preparation and cell lysis techniques.
Table 1 summarizes 35 publications, using samples derived from 17 different countries, evaluating the performance of amplification technologies in the clinical laboratory. The works in Table 1 are presented chronologically. Every effort has been made to accurately represent the original publication. There were, however, ambiguities in interpretation in some instances, and distinctive features of several papers. Clarifications are highlighted in the footnotes that follow.
Correlation of nucleic acid amplification results with culture results was chosen as the basis to compare the studies shown in Table 1. Using this perspective, this analysis reveals a conundrum: according to the methodologies outlined in Table 1, the sample subjected to amplification is derived, in the vast majority of instances, from the "button" used to seed the culture. Given the sensitivity of amplification relative to culture, a culture positive/amplification negative (e.g., false negative amplification) does not make intuitive sense. Several authors present "corrected correlations" (see Footnote C in Table 1). For example, if several false negatives were obtained, and the result could be resolved by further purification of the target DNA, dilution of inhibitors, multiple amplifications of the same sample, multiple amplifications of different samples from the same patient, or reamplification of the amplified specimen; the corrected results were presented.
This introduces an interesting dilemma. Jackson, J. B. et al., J. Clin. Micro. 31:3123-3128 (1993)) have successfully implemented an HIV-PCR quality assurance panel involving 11 laboratories. The reported sensitivity suggests that all laboratories have the routine capability of identifying 2 copies of the HIV genome in a background of 10.sup.6 human cells. The studies in Table 1 that are reviewed herein strongly suggest that the sensitivity of TB-amplification technologies is similar. What is apparent from the discussion of the studies in Table 1 is that, while the sensitivity of these amplification technologies for detection of Mycobacteria is expected to be orders of magnitude greater than that of either culture or smear, there are consistent aberrations to this expectation. While some of the anomalies are due to inhibition, and are, therefore, easily explained, many of the irregularities are "unexplained." What will become apparent is that, even if an internal control is used to detect inhibitors, these unexplained aberrations are reasonably common, and as such pose a significant obstacle to validation of amplification technologies for the detection of Mycobacteria in the clinical laboratory. For example, while positives, or suspected positives, may be rechecked, typically negatives are not. Since false negatives will occur with regularity, the laboratory will have no way of knowing which samples are truly false negative and which need to be "resolved." Consequently, patient care will be compromised as the diagnosis will incorrectly appear negative. Therefore, for the purpose of discussion of the published art herein, the art's uncorrected data was used. These aberrations--the false negatives--are the focus of the remaining discussion.
TABLE 1 __________________________________________________________________________ Art Reported Correlation of Culture Results with Nucleic Acids Amplification Results. For each study in Table 1 the processing protocol is abbreviated (see Footnote A), the number of clinical specimens evaluated is stated with the corresponding number of culture and amplification positive samples (see Footnote B), and then the "uncorrected" correlation to culture shown (see Footnote C). These 35 were chosen because a clear comparison between the amplification results and the culture results was presented. The studies of Shankar, P. et al., Lancet 337:5-7(1991); deWit, D. et al., Tubercle and Lung Dis. 73:262-267 (1992); and Kaneko, K. et al., Neurology 40:1617-1618 (1990) were omitted because no clear comparison on this basis could be made. .sup.B Total: Author(s) .sup.A Processing Protocol Cult.sym./Amp.sym. .sup.C Correl. __________________________________________________________________________ Brisson-Noel et al., Lancet ii: 1069-1071 (1989) SDS.sub.P NaOH/SDS/95.degree./15' Org/ppt 35: 13/15 100% Shankar et al., Lancet 335:423 (1990) NALC NaOH/NaCl/SDS/95.degree./15' Org/ppt .sup.D 23: 100% Hermans et al., J. Clin. Micro. 28:1204-1213 (1990) (NaOH/NaHPO.sub.4).sub.P (TE).sub.W Lz/SDS/PrK Org/ .sup.B 17: 100% ppt Pao et al., J. Clin. Micro. 28:1877-1880 (1990) OxAc NaCl/EDTA Lz Phenol/SDS 284: (100%).sup.F .sup..dagger-dbl. Sjobring et al., J. Clin. Micro. 28:2200-2204 DTT TE/95.degree. Sonic .sup.G 7: 75% DEWit et al., J. Clin. Micro. 28:2437-2441 (1990) PEG 70.degree./30' Phenol/SDS .sup.H 26: 100%4 Thierry et al., J. Clin. Micro. 28:2668-2673 (1990) () NaOH/SDS/95.degree./15' Org/ppt .sup.I 75: 100%5 .sup..dagger..dagger-dbl. Pierre et al., J. Clin. Micro. 29:712-717 (1991) (a) SDS.sub.P NaOH/SDS/95.degree./15' .sup.J 82: (79.2%).sup.K (b) SDS.sub.P (PBS).sub.W PCR.sub.Buf /PrK/NonI 95.degree. C./10' Del Portillo et al., J. Clin. Micro. 29:2163-2168 (1991) (Water/95.degree./10').sub.P Lz SDS/PrK Org/ppt .sup.L 30: 100%8 .sup..dagger..dagger-dbl..sctn. Brisson-Noel et al., Lancet 338:364-366 (1991) SDS.sub.P NaOH/SDS/95.degree./15' Org/ppt .sup.M 446: (89.4%).sup.N .sup..dagger-dbl. Sritharan et al., Mol. Cell. Probes 5:385-395 NALC.sub.P (TEX).sub.W TEX/95.degree./30' .sup.O 96: (95.9%).sup.P Eisenach et al., Am. Rev. Resp. Dis. 144:1160-1162 (1991) NALC.sub.P Lz NaOH/SDS/95.degree./5' GuSCN/Si 162: 97.7% Manjunath et al., Tubercle 72:21-27 (1991) NALC NaOH/SDS/95.degree./15' Org/ppt 117: 100%1 .sup..paragraph. Cousins et al., J. Clin. Micro. 30:255-258 NALC.sub.P (Water).sub.W Lz SDS/PrK TMA/Org/ppt 177: 98.4% .sup..dagger. VAN DER Giessen et al., J. Clin. Micro. 30:1216-1219 (a) NaOH.sub.S Phenol/CHCl.sub.3 Silica H.sub.2 O .sup.Q 87: 30/5 & 16%, 3% (1992) (b) NaOH.sub.S Phenol/CHCl.sub.3 Silica H.sub.2 O 87: 30/7 & 23%, 3% (c) IDEXX.sub.BUF Column Pellet NaOH 120.degree./10' 87: 30/4 & 13%, 13% .sup..dagger..dagger-dbl..paragraph. Buck et al., J. Clin. Micro. 30:1331-1334 (1992) NALC.sub.P 2x(Water).sub.W Sonic 95.degree./10' .sup.R 43; 92.0% .sup..dagger..dagger-dbl..sctn. Victor et al., J. Clin. Micro. 30:1514-1517 (1992) (a) NALC PCR.sub.Buf 95.degree./10' .sup.S 155: 76.3%00 (b) NALC Sucrose (PBS).sub.W 155: 98.5%31 .sup..dagger. Soini et al., J. Clin. Micro. 30:2025-2028 NaOH/SDS.sub.P NaOH/SDS/95.degree./15' Org/ppt 127: 55.9% Altamirano et al., J. Clin. Micro. 30:2173-2176 (1992) NALC.sub.P Lz/SDS perCl Org/ppt 200: 97.7% .sup..dagger. Fauville-Dufaux et al., Eur. J. Micro. Inf. SDS or TriPO.sub.4 NaOH/SDS/95.degree./15' Org/ppt .sup.T 206: (91.3%).sup.U 11:797-803 (1992) .sup..dagger..dagger-dbl. Kolk et al., J. Clin. Micro. 30:2567-2575 (1992) NALC.sub.P (TX).sub.W TX/PrK/60.degree./18 hrs 95.degree./15' 227: 95.6% .sup..dagger..dagger-dbl..sctn..paragraph. Shawar et al., J. Clin. Micro. 31:61-65 (1993) NALC.sub.P (TEX).sub.W TEX/95.degree./30' 384: 78.9% .sup..sctn..paragraph. Wilson et al., J. Clin. Micro. 31:776-782 (a) NaOH (PBS).sub.W PBS/80.degree./20' GuSCN/Si .sup.V 171: (66.7%).sup.W (b) NaOH (PBS).sub.W PBS/80.degree./20' CHCl.sub.3 171: (89.9%) .sup..dagger. Folgueira et al., J. Clin. Micro. 31:1019-1021 NALC.sub.P PCR.sub.Buf /PrK/NonI 95.degree./10' .sup.X 75: (90.7%).sup.Y Kocagoz et al., J. Clin. Micro. 31:1435-1438 (1993) NALC.sub.P 3x(TE).sub.W .sup.Z 78: (100%).sup..alph a. .sup..dagger..dagger-dbl..paragraph. Forbes et al., J. Clin. Micro. 31:1688-1694 (1993) NaOH PBS/NonI 95.degree./10' Sonic .sup..beta. 173: (80.6%).sup..gam ma. 727: (83.8%) .sup..dagger..paragraph. Nolte et al., J. Clin. Micro. 31:1777-1782 (1993) NALC.sub.P TEX 95.degree./30' 313: 91.1%13 .sup..dagger..dagger-dbl. Irula et al., J. Clin. Micro. 31:1811-1814 (1993) Fic/Hyp/PBMC (TEX).sub.W ChX/56.degree./30' 5.degree./30' .sup..delta. 243: 15/44 80.0% .sup..dagger..dagger-dbl..sctn..paragraph. Clarridge et al., J. Clin. Micro. NALC.sub.P (TEX).sub.W TEX/95.degree./30' 1,166: (83.0%).sup..eps ilon. 31:2049-2056 (1993) Miyazaki et al., J. Clin. Micro. 31:2228-2232 (1993) DTT Lz PrK/SDS Org/ppt 417: (96.4%).sup..zet a. .sup..dagger..dagger-dbl..sctn. Jonas et al., J. Clin. Micro. 31:2410-2416 (1993) NALC HCl Gen.sub.Buf Sonic .sup..eta. 758: 119/116 (79.8%).sup..the ta. .sup..dagger-dbl..sctn..paragraph. Abe et al., J. Clin. Micro. 31:3270-3274 (1993) (a) Gen: NALC PBS/HCl Gen.sub.Buf Sonic 95.degree./ .sup..iota. 135: (90.6%).sup..kap pa. 15' (b) PCR: NALC TEX/GuSCN/Si 135: (81.2%) .sup..dagger..dagger-dbl..paragraph. Miller et al., J. Clin. Micro. 32:393-397 (1994) (a) Gen: NALC HCl Gen.sub.Buf Sonic 95.degree./15' .sup..lambda. 750: 142/131 (83.9%).sup..mu. (b) PCR: (NALC).sub.P PCR.sub.Buf Sonic 95.degree./15' 156: (78.2%) .sup..dagger..sctn. Pfyffer et al., J. Clin. Micro. 32:918-923 (a) NALC HCl Gen.sub.Buf Sonic 95.degree./15' .sup..nu. 515: (92.9%).sup..xi. (b) SDS HCl Gen.sub.Buf Sonic 95.degree./15' 423: (97.2%) .sup..dagger-dbl. Bodmer et al., J. Clin. Micro. 32:1483-1487 SDS HCl Gen.sub.Buf Sonic 95.degree./15' .sup..pi. 617: 71.4% __________________________________________________________________________ (A) All studies, with the exception of Hermans, P.W.M. et al., J. Clin. Micro. 28:1204-1213 (1990), DEWit, D. et al., J. Clin. Micro. 28:2437-2441 (1990), Del Portillo, P. et al., J. Clin. Micro. 29:2163-2168 (1991), Irula, J.V. et al., J. Clin. Micro. 31:1811-1814 (1993), and possibly Thierry, D. et al., J. Clin. Micro. 28:2668-2673 (1990), process the specimens for culture first. In general, all specimen processing techniques utilized are summarized in Kent, P.T. et al., "Public Health Mycobacteriology" in A Guide for the Level III Laboratory, U.S. Department of Health and Human Services, Centers for Disease Control, 1985, pp. 31-46. The symbology used to describe the processing of clinical specimens is as follows: "NALC" refers to the N-Acetyl-L-Cysteine/sodium hydroxide liquification/decontamination protocol described by Kubica, G.P.W. et al., Am. Rev. Resir. Dis. 87-775- 779 (1963). "DTT" refers to utilization of dithiothreitol (also known as Sputolysin .TM.) for the processing of Mycobacteria according to the protocol of Hirsch, S.R. et al., J. Lab. Clin. Med. 74:346-353 (1969). "SDS" (sodium dodecylsulfate) refers to the protocol of Engbaek, H.C. et al., Scand. J. Respir. Dis. 48:268-284 (1967). "Fic/Hyp" refers to the Ficoll/Hypaque gradient protocol for the purification of peripheral blood mononuclear cells (PBMC) after the method of Boyum (Boyum, A.J. Clin. Lab. Invest. 21(Suppl. 97):77-109 (1968)). "NaOH" refers to sodium hydroxide decontamination protocols specific to the authors, or is a direct quote from the author where the specifics of the decontamination protocol are not clearly stated. "OxAc" refers to the use of oxalic acid (Corper, H.J. et al., J. Lab. Clin. Med. 15:348-369 (1930). "PEG" refers to the use of polyethylene glycol precipitation specific to the authors. "TriPO.sub.4 " refers to the triphosphate procedure of Collins et al., Organization and Practice in Tuberculosis Bacteriology, Buttersworth, London, 1985. Invariably, the processed specimen is centrifuged to produce a "button" (sediment). The button is resuspended in either water or phosphate buffered saline (PBS) and aliquots removed for culture and smear. The remaining sediment is then further processed for amplification. A subscript "P" attached to the processing designation refers to the fact that the resuspended sediment was again centrifuged prior to further processing for amplification, and it is the pelleted sediment that was used in the study. A subscript "S" attached to the processing designation refers to the fact that the supernatant clarified by centrifugation, and it was the supernatant that was used in the study. Following processing for culture, several authors compare different protocols for preparation of the sample for amplification: only the protocol used to prepare specimens for the actual study are included. Those studies that compare sample preparation methodologies are identified in the text and the results of these findings discussed as well. The following buffer abbreviations apply: "TE" = Tris/EDTA, "TX" = Tris/Triton-X100, "TEX" = Tris/EDTA/Triton-X100, and "NonI" = Tris/Tween20 & NP-40. A subscript "W" refers to a wash step using the parenthetic buffer. "Gen.sub.Buf," "IDEXX.sub.Buf " and "PCR.sub.Buf " refer to suspension/resuspension of the sample in Gen-Probe buffer or IDEXX buffer (the compositions of which are unknown) or PCR buffer (20 mM Tris-HCl pH8.3, 50 mM KCl, 2.5 mM MgCl.sub.2, 0.45% Tween20, 0.45% NP-40), respectively. "CHCl.sub.3," "HCl," and "perCl" refer to inclusion of chloroform, hydrochloric acid and perchloric acid, respectively, in the processing steps. "PrK" and "Lz" refer to enzymatic digestion of the sample with proteinase K and/or lysozyme. Any step involving boiling or treatment at extreme temperature is simplified as 95.degree. for the prescribed length of time. "Org/ppt" refers to any extraction methodology involving organic compounds, or combinations thereof, including phenol, chloroform, and/or isoamyl alcohol, followed by precipitation of the nucleic acid with either ethanol or isopropanol. There are numerous variations on this theme, but all methodologies are similar to that found in Maniatis, T. et al., "Molecular Cloning A Laboratory Manual," Cold Spring Harbor Laboratory, New York, 1982, pp. 458-463). "TMA" refers to the utilization of quaternary amines to precipitate DNA or polysaccharides similar to the methodology of Baess, I. Acta Path. Microbial. Scand. Sect. B, 82:780-784 (1974). "GuSCN/Si" refers to purification of DNA from the sample using guanidinium isothiocyanate in the presence of silica, as described, for example, by Boom, R. et al., J. Clin. Microbial. 28:495-503 (1990). "Sonic" refers to release of the DNA by sonication of the sample (in the presence or absence of glass beads). (B) The total number of clinical samples used in a given study is followed first by the number of culture positives and then by the number of amplification positives. Additional footnotes referred to in this column describe pertinent characteristics of the study. (C) Shown is the uncorrected correlation between the amplification result and the culture result (e.g., the number of amplification positive specimens that were also culture positive, divided by the total number of culture positives). Several authors have further examined discrepant results (e.g., "false negatives" or "false positives"), or incorporated patient information into the analysis and published corrected results. The correlation listed in this column refers to uncorrected results. If a discrepancy between the published result and the result listed in this column exists, the number is in parentheses with an appropriate footnote explaining the difference. (D) Shanker, P. et al., Lancet 335:423 (1990) do not explicitly state the protocol used to process samples for either culture or PCR. However, Manjunath, N. et al., Tubercle 72:21-27 (1991) state that their publication is the final version of preliminary work presented in the Shanker, P. et al., Lancet 335:423 (1990) study. Therefore, it is assumed, for the purpose of discussion, that the processing conditions are identical to those of Manjunath, N. et al., Tubercle 72:21-27 (1991). (E) The study of Hermans, P.W.M. et al., J. Clin. Micro. 28:1204-1213 (1990)) process crude specimens directly for amplification (e.g., the specimen was not processed for culture and the sediment split for analysis by both culture and PCR). (F) Pao, C.C. et al., J. Clin. Micro. 28:1877-1880 (1990) do not give uncorrected results for PCR. They do, however, state that: "If the initial result was negative, a portion (usually one-fifth) of the first amplified reaction mixture was amplified for another 32 cycles . . ." Therefore, there may originally have been false negatives. (G) The slant corresponding to this false negative PCR result in the Sjobring et al., J. Clin. Micro. 28:2200-2204 (1990) study produced a single colony. The remaining three culture/PCR positive specimens were processed directly for amplification. (H) The study of DEWit, D. et al., Tubercle and Lung Dis. 73:262-267 (1992) uses pleural fluid aspirates exclusively. The crude specimens were precipitated by PEG directly and subjected to amplification. (I) Thierry, D. et al., J. Clin. Micro. 28:2668-2673 (1990) state: "DNA was extracted from 0.2 to 1 ml of crude or sodium hydroxide decontaminated clinical specimens . . ." Therefore, for the purpose of discussion, it is assumed that the specimens may have been used directly for amplification. (J) Pierre, C. et al., J. Clin. Micro. 29:712-717 (1991) discuss two methodologies for processing of clinical specimens. They do not explicitly state which is used for the study. They do, however, suggest that "In some cases, a portion of the sample . . ." was processed using the second, nonionic lysis methodology. They also state that more inhibitors were found in these samples. Therefore, it is assumed, for the purpose of discussion, that organic extraction was the primary protocol used in this study. (K) Pierre, C. et al., J. Clin. Micro. 29:712-717 (1991) utilize a "reamplification" protocol in which the tubes are opened, a portion is transferred to a fresh PCR mix and re-amplified. Their results were 79.2% using the initial amplification and 100% upon reamplification. (L) The study of Del Portillo, P. et al., J. Clin. Micro. 29:2163-2168 (1991) processed the sputum directly for amplification. (M) Brisson-Noel, A. et al., Lancet 338:364-366 (1991) test 514 specimens but only 446 were processed for both PCR and culture. (N) The system used by Brisson-Noel, A. et al., Lancet 338:364-366 (1991) is designed against the 65Kd protein and amplifies most Mycobacteria. Of 141 culture positive specimens, 130 were MTB and 11 were MOTT (Mycobacteria other than tuberculosis). Of the 141, 126 were positive by PCR, 6 were false negatives and 9 contained inhibitors. Therefore, the correlation to culture was 89.4% (126 .div. 141 = 0.894). The correlation reported is 97.4% and includes both elimination of samples with inhibitors and analysis of patient data (e.g., high clinical suspicion of samples that were PCR positive/culture negative). (O) Sritharan, V. et al., Mol. Cell. Probes 5:385-395 (1991) actually compare 8 different methodologies for the release of DNA for PCR. The boiling method was that chosen for the study. (P) The correlation stated by Sritharan, V. et al., Mol. Cell. Probes 5:385-395 (1991) is 100%, and represents reamplification of 3 samples that were false negative on the first PCR. The uncorrected correlation is 96.0%. (Q) van der Giessen, J.W.B. et al., J. Clin. Microbiol 30:1216-1219 (1992) compared three PCR based systems (McFadden, J.J. et al., Mol. Microbial. 1:283-291 (1987); van der Giessen, J.W.B. et al., J. Med. Microbiol. 36:255-263 (1992); Vary, P.H. et al., J. Clin. Microbial. 28:933-937 (1990)) designed to detect M. paratuberculosis in bovine feces. One of these is the commercially available kit from IDEXX (Vary, P.H. et al., J. Clin. Microbial. 28:933-937 (1990)). In this study they tested 87 samples by all three methods in two separate runs. The cultures were run only once (culure.sym. = 30), whereas each PCR was run twice, hence the two numbers. (R) Buck, G.E. et al., J. Clin. Micro. 30:1331-1334 (1992) actually compare 4 different methodologies for the release of DNA for PCR. The sonication method shown was that chosen for the study. In addition, all samples in the study were known TB-positives, and all were smear positive. (S) Victor et al., J. Clin. Micro. 30:1514-1517 (1992) split all sediments and either amplified directly, or further purified the sample by sucrose gradient fractionation. The results from both methodologies were available for discussion. (T) Fauville-Dufaux et al., Eur. J. Micro. Inf. Dis. 11:797-803 (1992) state that specimens were proeessed by " . . . either . . . " SDS or triphosphate. They do not differentiate further which samples were processed by which method. (U) The primer pair used by Fauville-Dufaux et al., Eur. J. Micro. Inf. Dis. 11:797-803 (1992) was designed against the antigen 85 sequence. Consequently, their system was able to amplify most Mycobacterial species. Of the 206 samples processed, 92 were culture positive. Of these 92, 84 were initially positive by PCR (91.3%). Of the 92, 82 were identified as MTB. Of these, 82, only 74 were initially positive by PCR (90.2%). Whereas all 8 missed were seen to contain inhibitors, 3 could be diluted to the point where a positive signal was observed. Therefore, upon resolution, the authors state that 77 of the 82 (93.9%) were correctly identified by PCR. The authors report their correlation as 93.9%. (V) Wilson, S.M. et al., J. Clin. Micro. 31:776-782 (1993) compare 2 different methodologies for the release of DNA for PCR. Both methods were used on all samples in this study, hence the two correlations. In addition, Wilson, S.M. et al., J. Clin. Micro. 31:776-782 (1993) use a "one-tube" nested protocol. (W) Wilson. S.M. et al., J. Clin. Micro. 31:776-782 (1993) state the correlation as 75% for the GuSCN/Si procedure and 92% for the chloroform procedure. This represents "patient results." The stated correlations in this column represent the published correlations of "individual specimen" results. In addition, all samples in this study were run in duplicate and discrepants resolved. (X) Folgueira, L. et al., J. Clin. Micro. 31:1019-1021 (1993) compare 2 different methodologies for the release of DNA for PCR. The boiling method shown was that chosen for the study. All specimens in this study were known TB-positives. (Y) The correlation between culture and PCR stated by Folgueira, L. et al., J. Clin. Micro. 31:1019-1021 (1993) is 100%, and represents inclusion of seven samples that required dilution of inhibitors. The uncorrected correlation is 90.7%. (Z) Kocagoz, T. et al., J. Clin. Micro. 31:1435-1438 (1993) compare two different methodologies for preparing the sediment for PCR. The boiling method shown was that chosen for the study. (.alpha.) The "gold standard" used by Kocagoz, T. et al., J. Clin. Micro. 31:1435-1438 (1993) is "high clinical suspicion." While the stated correlation to this criterion is given as 87%, the correlation to culture is 100%. (.beta.) Forbes, B.A. et al., J. Clin. Micro. 31:1688-1694 (1993) initially test 173 samples to optimize PCR. They then tested 727 samples. Information on both studies was available for discussion. (.gamma.) Forbes, B.A. et al., J. Clin. Micro. 31:1688-1694 (1993) state the correlation for the 727 samples examined as 87.2%. This sensitivity includes incorporation of patient data in analyzing PCR positive/culture negative samples. Table 2 of this work shows that of 80 MTB culture positives, 67 were picked up by PCR. Therefore, the uncorrected sensitivity would be 83.8%. (.delta.) The study of Irula, J.V. et al., J. Clin. Micro. 31:1811-1814 (1993) uses blood samples exclusively. In addition, the aim of this study is the identification of M. avium, as opposed to MTB. (.epsilon.) Clarridge, J.E. et al., J. Clin. Micro. 31:2049-2056 (1993) state several numbers for culture correlation. The actual number of culture positive MTB specimens is stated as 218, and the number of PCR positives within this group is stated as 181 (see Table 2 of this reference). Therefore, the uncorrected correlation recorded for discussion is 83.0%. (.zeta.) Miyazaki, Y. et al., J. Clin. Micro. 31:2228-2232 (1993) use a nested PCR system (e.g., sample from the first reaction was actually transferred to fresh PCR mix and reamplified) and identify 54 of the 56 that were culture positive (96.4% correlation). However, there were 10 samples that were culture negative/smear positive/PCR positive. The authors include these 10 and state the correlation as 97.0% (64 .div. 66 = 0.970). (.eta.) The study of Jonas, V. et al., J. Clin. Micro. 31:2410-2416 (1993) utilized the proprietary amplification scheme commercialized by Gen-Probe, exclusively. (.theta.) In the study of Jonas, V. et al., J. Clin. Micro. 31:2410-2416 (1993), of 119 culture positive specimens, only 95 were picked up by amplification (79.8%). There were 21 specimens that were culture negative/amplification positive. The authors concluded that only 17 of these 21 were true positives. Inclusion of these 17 suggests that amplification identified 112 of 136 true positives. This increases the sensitivity to that reported: 82.4%. (.iota.) In the study of Abe, C. et al., J. Clin. Micro. 31:3270-3274 (1993) all specimens were split such that the results derived from the Gen-Probe method (Gen), could be compared with the results derived from the PCR system described by Kolk, A.H.J. et al., J. Clin. Micro. 30:2567-2575 (1992). (.kappa.) Abe, C. et al., J. Clin. Micro. 31:3270-3274 (1993) resolve several discrepant results and state that the correlation of the Gen-Probe method is 91.9%, and 84.2% by PCR. Disregarding resolution of the discrepants gives 90.6% and 81.3%, respectively. The sensitivity of Gen-Probe amplification appears better than that obtained by PCR. (.lambda.) The principle amplification scheme used by Miller et al., J. Clin. Micro. 32:393-397 (1994) is that being commercialized by Gen-Probe. They do, however, compare these results with the results obtained by the PCR system of Eisenach et al., J. Clin. Micro. 26:2240-2245 (1988). Whereas all 750 samples were amplified by the Gen-Probe methodology, only those specimens considered positive (156) were amplified by PCR. (.mu.) 156 specimens in the Miller et al., J. Clin. Micro. 32:393-397 (1994) study were " . . . culture-proven and/or clinically diagnosed [with] tuberculosis." Of these 156, only 142 were actually culture positive. Using the Gen-Probe scheme, 131 of the 156 (83.9%) were positive. Subsequently, "[r]epeat testing of specimens with a different aliquot from the same processed sediment increased detection to 142 of 156 (91.0%)." The actual data is presented in a way that makes it all but impossible to determine if this number represents the correlation to culture. Amplification of the 156 positives by PCR initially identified 122 (78.2%) specimens. 143 (92.3%) were identified upon repeated PCR testing (the sensitivity of PCR appears only marginally better than that of the Gen-Probe method). There were a number of smear positive specimens that were missed on initial examination by both methodologies. All smear positives were identified on repeat testing. (.nu.) Pfyffer et al., J. Clin. Micro. 32:918-923 (1994) produced sediments by two different methods, NALC and SDS, and amplified all sediments using the Gen-Probe methodology. Specimens were unique for each category (e.g., the same specimen was processed by one or the other of the methods, not both). (.xi.) In the Pfyffer et al., J. Clin. Micro. 32:918-923 (1994) study, 42 NALC specimens were culture positive, and 39 (92.9%) were identified by amplification. Thirty-six SDS specimens were culture positive and 35 (97.2%) were identified by amplification. However, the authors resolve amplification positive/culture negative results and conclude that amplification of NALC sediments correctly identified 46 of 49 "true positives" (93.9%), and amplification of SDS sediments correctly identified 38 of 39 "true positives" (97.4%). (.pi.) Bodmer et al., J. Clin. Micro. 32:1483-1487 (1994) used the Gen-Probe amplification scheme. (.dagger.) Indicates those studies in which the authors discuss inhibition or inhibitors as a contributing factor to false negatives. (.dagger-dbl.) Indicates those studies in which the authors discuss low copy numbers as a contributing factor to false negatives. (.sctn.) Indicates those studies in which the authors discuss the phenomenon of statistical dropouts, or a phenomenon that is consistent with statistical dropouts (this would include "unexplained" results), or discuss the possibility of aberrant being exacerbated by clumping. (.paragraph.) Indicates those studies in which examples of culture positive/smear positive/amplification negative specimens were observed.
D. The Studies of Table 1
Examination of the data reported in the publications of Table 1 suggests that, regardless of system design or sample processing technique, there was a wide variation in results. These studies include literally all possible sources of specimen; including sputum, bronchial washes, pleural fluid, gastric aspirates, cerebrospinal fluid (CSF), urine, tissue biopsy, bone marrow, abscess and exudates, blood, serum, peritoneal fluid and feces. Two different amplification schemes were used: 30 studies used PCR exclusively, three studies used the isothermal, retroviral type, proprietary amplification scheme being commercialized by Gen-Probe (Jonas, V. et al., J. Clin. Micro. 31:2410-2416 (1993); Pfyffer et al., J. Clin. Micro. 32:918-923 (1994); Bodmer et al., J. Clin. Micro. 32:1483-1487 (1994)), and two studies compared the two amplification technologies (Abe, C. et al., J. Clin. Micro. 31:3270-3274 (1993); Miller et al., J. Clin. Micro. 32:393-397 (1994)). Thirty-three studies focussed on detection of MTB, one study focussed on M. avium diagnosis (Irula, J. V. et al., J. Clin. Micro. 31:1811-1814 (1993)), and one study pursued M. paratuberculosis detection (van der Giessen, J. W. B. et al., J. Clin. Microbiol. 30:1216-1219 (1992)). False negative results could be found in most specimen categories, regardless of target, processing technique or amplification technology.
The studies reported in Table 1 range in sample size from 7 to 1,166 specimens. Correlations with culture ranged from 3% to 100%. Nine of the 35 studies (26%) claim correlations of 100%. However, the majority, 7 of 9 (78%), involve sample sizes of less than 100 (n&lt;100). Only two studies (Manjunath, N. et al., Tubercle 72:21-27 (1991); Pao, C. C. et al., J. Clin. Micro. 28:1877-1880 (1990)) used more than 100 specimens (however, Pao, C. C. et al., J. Clin. Micro. 28:1877-1880 (1990) may have reamplified in an effort to confirm negative samples: see footnote F in Table 1). Alternatively, 26 of the 35 studies (74%) show correlations of less than 100%. In this group, 20 of the 26 (77%) utilize sample sizes greater than 100 (n&gt;100). It appears that there is an inverse relationship between the amplification-culture correlation and sample size: in general, the more samples included in a study, the lower the correlation.
In Table 1, 32 of the 35 studies (91%) show that amplification was able to detect the presence of Mycobacterial DNA in culture negative specimens (two of the three remaining studies used known positive specimens only). Two of these 32 more than double the number of culture positives (Irula, J. V. et al., J. Clin. Micro. 31:1811-1814 (1993); Pao, C. C. et al., J. Clin. Micro. 28:1877-1880 (1990)), and three very nearly double this number (Kolk, A. H. J. et al., J. Clin. Micro. 30:2567-2575 (1992); Manjunath, N. et al., Tubercle 72:21-27 (1991); Miyazaki, Y. et al., J. Clin. Micro. 31:2228-2232 (1993)). Thirty-one authors state directly, or reference the fact, that under ideal in vitro conditions their systems have the ability to detect the presence of 10 copies or less. The sensitivity of the remaining 4 range from 15 to 40 copies (Altamirano, M. et al., J. Clin. Micro. 30:2173-2176 (1992); Hermans, P. W. M. et al., J. Clin. Micro. 28:1204-1213 (1990); Pao, C. C. et al., J. Clin. Micro. 28:1877-1880 (1990); Soini, H. et al., J. Clin. Micro. 30:2025-2028 (1992)). Culture negative samples, that are positive by amplification, are more easily explained than culture positive-amplification negative samples: contrary to culture, PCR does not require viable organisms. For example, processing is known to kill the organisms, drug therapy may have already compromised viability, or low copy number combined with reduced viability might all contribute to the former class of samples. Regardless of system parameters, amplification should have superior sensitivity relative to either culture or smear.
Among the 35 studies, 8 different methods are used to process the raw specimens. Seventeen studies processed samples for both culture and amplification by treatment with N-acetyl-L-cysteine/NaOH (NALC), 6 studies used sodium dodecyl sulfate (SDS), 3 studies used sodium hydroxide (NaOH), 2 studies used dithiothreitol (DTT); and oxalic acid (OxAc), polyethylene glycol (PEG), and the ficoll-hypaque (Fic/Hyp) gradient method were each used once. One study actually compared processing by NALC with that of SDS (Pfyffer et al., J. Clin. Micro. 32:918-923 (1994)). Three studies avoided using the culture sediment and directly processed samples for amplification. Of the nine studies that obtained correlations of 100%, three used NALC, three processed specimens for amplification directly, one study used SDS, one used OxAc, and one used PEG. Keeping in mind that 7 of these studies utilized sample sizes of less than 100 (n&lt;100), no conclusions can be drawn between how the raw specimens were processed, and perfect correlation. The study of Pfyffer et al., J. Clin. Micro. 32:918-923 (1994) compared NALC and SDS processing and concluded that neither method was superior.
Preparation of the processed sediment (or samples directly) for amplification falls into seven basic categories: (i) sixteen examples use variations on organic extraction and alcohol precipitation methodologies (org/ppt) as described by Maniatis, T. et al. ("Molecular Cloning A Laboratory Manual," Cold Spring Harbor Laboratory, New York (1982), pp. 458-463); (ii) two examples use an enzymatic lysis and boiling protocol (Lz/Prk.fwdarw.95.degree./15') as described by Higuchi, R. Amplifications 2:1-3 (1989), and six examples simply boil the specimen; (iii) ten examples use sonication (sonic); (iv) three use chaotropic agents and glass beads (GuSCN/Si) as described by Boom, R. et al., J. Clin. Microbiol. 28:495-503 (1990), and two use a similar protocol of binding the DNA to silica; (v) one example used sucrose gradient fractionation (Victor, T. et al., J. Clin. Micro. 30:1514-1517 (1992)); (vi) one example used Chelex-100 as described by de Lamballerie, X. et al., Res. Microbiol. 143:785-790 (1992); and (vii) one example (van der Giessen, J. W. B. et al., J. Clin. Microbiol. 30:1216-1219 (1992)) used the column chromatography procedure suggested by the manufacturer (Vary, P. H. et al., J. Clin. Microbiol. 28:933-937 (1990)). The sum is greater than 35 because the studies of van der Giessen, J. W. B. et al., J. Clin. Microbiol. 30:1216-1219 (1992), Victor, T. et al., J. Clin. Micro. 30:1514-1517 (1992), Wilson, S. M. et al., J. Clin. Micro. 31:776-782 (1993) and Abe, C. et al., J. Clin. Micro. 31:3270-3274 (1993), Miller et al., J. Clin. Micro. 32:393-397 (1994) and Pfyffer et al., J. Clin. Micro. 32:918-923 (1994) process all specimens by more than one protocol and present analyses for each method. Eight of the nine studies that claim 100% correlation isolate the DNA by organic extraction/alcohol precipitation methodologies. However, eight studies that also use organic extraction methodologies had correlations ranging from 55.9% to 98.4%. One study that claimed 100% correlation was processed by boiling (Kocagoz, T. et al., J. Clin. Micro. 31:1435-1438 (1993)). Alternatively, six studies using this same protocol reported correlations between 78.9% and 95.9%. Seven studies actually compare methods to prepare the culture sediment for amplification. Their conclusions differ as follows: Pierre, C. et al., J. Clin. Micro. 29:712-717 (1991)) selected organic extraction/alcohol precipitation; Wilson, S. M. et al., J. Clin. Micro. 31:776-782 (1993)) simply treated with chloroform; Folgueira, L. et al., J. Clin. Micro. 31:1019-1021 (1993) preferred the enzymatic lysis/boiling method; Kocagoz, T. et al., J. Clin. Micro. 31:1435-1438 (1993) and Sritharan, V. et al., Mol. Cell. Probes 5:385-395 (1991) chose the simple boiling method; and Forbes, B. A. et al., J. Clin. Micro. 31:1688-1694 (1993) and Buck, G. E. et al., J. Clin. Micro. 30:1331-1334 (1992) identified sonication as the optimal method (only the study of Kocagoz, T. et al., J. Clin. Micro. 31:1435-1438 (1993) achieved 100% correlation). It would appear that the occurrence of false negatives is not only independent of the protocol employed to prepare the sample for amplification, but there is controversy surrounding this issue as well.
Two studies actually compare amplification methods: Abe, C. et al., J. Clin. Micro. 31:3270-3274 (1993) show that the Gen-Probe method was marginally better than PCR (see footnote .kappa. in Table 1), while Miller et al., J. Clin. Micro. 32:393-397 (1994) determine the opposite (see footnote .mu. in Table 1). Apparently, neither amplification technique confers a significant advantage for clinical diagnosis of TB infections.
E. PCR Inhibitors
Of the papers reporting correlations less than 100%, 17 studies refer to amplification "inhibitors" as a contributing factor to false negatives (see those authors with a superscript .dagger. in Table 1). Blood (Panaccio, M. et al., Nucl. Acids Res. 19:1151 (1991)), feces (Allard, A. et al., J. Clin. Microbiol. 28:2659-2667 (1990)), sputum (Hermans, P. W. M. et al., J. Clin. Micro. 28:1204-1213 (1990); Shawar, R. M. et al., J. Clin. Micro. 31:61-65 (1993)) and urine (Khan, G. et al., J. Clin. Pathol. 44:360-365 (1991)) all contain PCR inhibitors. In addition, with respect to sputum, bronchial washes and tracheal aspirates, there is a direct correlation between the viscosity of the specimen (mucous content) and disease state: patients with advanced stages of tuberculosis have the most viscous sputum and these specimens have the highest probability of retaining amplification inhibitors. Hermans, P. W. M. et al., J. Clin. Micro. 28:1204-1213 (1990)) and Shawar, R. M. et al., J. Clin. Micro. 31:61-65 (1993)) show reductions in sensitivity of 5-20 fold, and 5 fold, respectively, in the presence of sputum.
Of the 42 methodologies presented for processing in Table 1 only twelve do not incorporate some form of buffer exchange. For example, organic extraction/precipitation, washing of the pellet, or protocols using chaotropic agents (GuSCN/Si), all require a buffer exchange at some point. Sonication of the sediment, however, does not require a buffer exchange. None of these twelve studies achieves 100% correlation, and nine within this group refer to inhibitors as a contributing factor to false negatives. Inhibitors appear to be derived from both the specimen and solutions used for processing, and both sources pose significant challenges to clinical implementation of amplification technologies.
F. Low Copy Numbers, Statistical Dropouts and "Unexplained" Results
Statistical dropouts, also referred to as "sample bias," are due to low copy number; in a sample with extremely low copy numbers, from which aliquots must be taken, there exists the possibility that some aliquots will contain no target. For example, if there are ten copies of the target in a milliliter, and ten 100 .mu.l aliquots are taken, target will be absent from some fractions. These aliquots, while being interpreted as false negatives, are "true amplification negatives." Eight studies in Table 1 describe a phenomenon that could be explained by this type of sample bias (see those authors' names with a superscript .sctn. in Table 1). As discussed below, this phenomenon is greatly exacerbated by aggregation.
Of the papers reporting correlations less than 100%, 15 refer to "low copy number" directly as a contributing factor to false negative results (see those authors with a superscript .dagger-dbl. in Table 1). However, 6 of these 15, plus 3 others, present examples where negative amplification specimens were both culture positive and smear positive (see those authors' names with a superscript .paragraph. in Table 1). The limit of detection of acid fast staining has been reported as 7,800 to 9,500 organisms per milliliter of sputum (Hobby, G. L. et al., Antimicrob. Ag. Chemother. 4:94-104 (1973); Yeager, H. et al., Amer. Rev. Resp. Dis. 95:998-1004 (1967)). Clarridge, J. E. et al., J. Clin. Micro. 31:2049-2056 (1993) present an extensive analysis of false negatives (see Table 7 of this reference). Of 37 false negative specimens analyzed in detail, 26 showed dropouts, while 11 were "true" PCR negatives. Nine of these 37 were smear positive: 4 of these 9 contained inhibitors, 3 were not tested for inhibitors, and 2 were found to be free of inhibitors. Of these last two, one was a true PCR negative. Shawar, R. M. et al., J. Clin. Micro. 31:61-65 (1993) also reported culture positive-smear positive-PCR negative specimens that were seen to be free of inhibitors. If the sample does not contain inhibitors, and is smear/culture positive, "low copy number" cannot be a realistic possibility. Shawar, R. M. et al., J. Clin. Micro. 31:61-65 (1993) refer to this class of false negatives as "unexplained."
G. Partitioning of Mycobacteria During Centrifugation
The buoyant nature of Mycobacterium was evident as early as 1924 (Andros, P. M. et al., Am. Rev. Tuberc. 9:99 (1924)). Since then, several studies have highlighted the difficulty of sedimenting Mycobacteria (Hanks, J. H. et al., J. Lab. Clin. Med. 23:736-746 (1938); Hata, Jr., D. et al., Dis. Chest 18:352-362 (1950); Klein, G. C. et al., Am. J. Clin. Pathol. 22:581-585 (1952); Ratman, S. et al., J. Clin. Microbiol. 23:582-585 (1986); Rickman, T. W. et al., J. Clin. Microbiol. 11:618-620 (1980); and Robinson, L. et al., J. Lab. Clin. Med. 27:84-91 (1941)), and, in several instances, culturing the supernatant is advocated as standard practice.
While several studies report that the supernatant fractions contained smear positive material (Hanks, J. H. et al., J. Lab. Clin. Med. 23:736-746 (1938); Rickman, T. W. et al., J. Clin. Microbiol. 11:618-620 (1980)), another study showed that in 88.8% and 82.4% of all specimens centrifuged at 2,000 rpm and 3,000 rpm, respectively, the supernatant was culture positive (Klein, G. C. et al., Am. J. Clin. Pathol. 22:581-585 (1952)). In fact, this same study showed that in 2.2% and 2.7% of all specimens centrifuged at 2,000 rpm and 3,000 rpm, respectively, the sediment was culture negative while the supernatant was culture positive. Analyzing the supernatant fraction is still discussed in contemporary laboratory manuals (Kent, P. T. et al., "Public Health Mycobacteriology" in A Guide for the Level III Laboratory, U.S. Department of Health and Human Service, Centers for Disease Control, (1985) pp. 31-46; Sommers, H. M. et al., "Mycobacterium," in: Manual of Clinical Microbiology, E. H. Lennette et al., eds., 4th ed., Am. Soc. Microbiol., Washington, D.C. (1985), pp. 216-248).
The inverse relationship between sample size and correlation could potentially be explained by the buoyancy phenomenon. Larger sample sizes require batch processing. During batch processing the time it takes between work-up of the first and last specimens increases. As this time increases, buoyancy has a greater amount of time to exert an effect. It has been suggested that the source of this buoyancy is the high lipid content of these organisms (Silverstolpe, L. Nord. Med. 40/48:2220-2222 (1948)).
H. Influence of the Cell Wall and Surface Tension on Recovery of Mycobacteria
The nature of the cell wall of the Mycobacteria is responsible for their survival tenacity. Micrographs reveal a very complex structure 30-40 nm thick (Rastogi, N. et al., Antimicrob. Agents Chermother. 20:666-677 (1981)). As much as 60% of the dry weight of the cell wall is lipid (Joklik, W. K. et al., Zinsser Microbiology 20th edition, Appleton & Lange, Norwalk, Conn. (1992), pp. 499).
The cell wall of the Mycobacterium has three distinct layers: (i) the peptidoglycan, (ii) the arabinogalactan, and (iii) glycolipids (for a comprehensive review of cell wall structure see McNeil, M. R. et al., Res. Microbiol. 142:451-463 (1991)). Mycolic acids, which are extremely hydrophobic and consist primarily of hydrocarbon chains (.SIGMA.=C.sub.76 -C.sub.80), are used extensively in the construction of both the arabinogalactan and the glycolipid layers. The structure and species distribution of mycolic acids is reviewed in Takayama, K. et al., "Structure and Synthesis of Lipids," in: The Mycobacteria: a Source Book, Part A, G. P. Kubica et al., eds., Marcel Dekker, Inc., New York, N.Y. (1984), pp. 315-344. The Mycobacteria are essentially encased in wax.
M. tuberculosis form "cords" during growth (cords are clumps or aggregates of large numbers of organisms), and there is a direct relationship between MTB virulence and cord formation (Joklik, W. K. et al., Zinsser Microbiology 20th ed., Appleton & Lange, Norwalk, Conn. (1992), pp. 503). Organisms of the MAC complex have additional glycolipid (C-mycoside) components in their cell wall (the differences among the serovars is summarized in Brennan, P. J. Rev. Infect. Dis. 11 (Supp. 2): s420-s430 (1989)). In culture, while MTB complex organisms are seen to form dense clumps, M. avium grows in a more diffuse, single cell fashion (Dubos, R. J. et al., J. Exp. Med. 83:409-423 (1946)).
Dubos, R. J. Proc. Soc. Exp. Biol. Med. 58:361-362 (1945) observed that the hallmark pellicle growth of cultured MTB could be modified by the addition of the polyoxyalkylene derivative of sorbitan monostearate (Tween 60; CAS.RTM.No. 9005-67-8). This observation was later extended to show that other similar derivatives could cause MTB to exhibit "rapid," "diffuse" and "submerged" growth characteristics (Dubos, R. J. et al., J. Exp. Med. 83:409-423 (1946)). Tween 80 (CAS.RTM.No. 9005-65-6) was found to be the most active in that regard. These authors concluded that submerged growth was due to "wetting" of the cell surface. The term "wetting" is used exclusively in the context of surface tension. The implication of these studies was that pellicle growth resulted from surface tension between the waxy coat and the aqueous media, and that the addition of Tween 80 alleviated this physical interaction.
If surface tension confined the organisms to the surface, this, in combination with cording, could explain aberrant results: very sick patients are infected by organisms that have a propensity for cord formation. In addition, large cords produce smear positive results and cultures that turn positive fairly quickly. Large cords would also exacerbate sample bias because a large cord would distribute as a single copy, but have the potency of thousands of copies. Consequently, as suggested by Klein, G. C. et al., Am. J. Clin. Pathol. 22:581-585 (1952), the microorganisms would easily be poured off with supernatant fraction, thereby facilitating the sample bias phenomenon. In addition, cording would cause the bacteria to partition such that smear/culture positive and amplification negative results would occur.
If surface tension and aggregation were responsible for the anomalous results, and surface tension and aggregation could be overcome by the addition of nonionic detergents, then it seems logical that these reagents should improve correlation to culture: five studies in Table 1 use nonionic detergents to wash sediments prior to amplification (Clarridge, J. E. et al., J. Clin. Micro. 31:2049-2056 (1993); Irula, J. V. et al., J. Clin. Micro. 31:1811-1814 (1993); Kolk, A. H. J. et al., J. Clin. Micro. 30:2567-2575 (1992); Shawar, R. M. et al., J. Clin. Micro. 31:61-65 (1993); and Sritharan, V. et al., Mol. Cell. Probes 5:385-395 (1991)). The correlations to culture range from 78.9% to 95.5% in this subset. As early as 1941 it was recognized that agents that alleviated surface tension were impotent in enhancing recovery by centrifugation (Robinson, L. et al., J. Lab. Clin. Med. 27:84-91 (1941)). Therefore, the art teaches there clearly is no additional advantage to inclusion of these detergents in the wash buffer.
I. Problems Unique to MAC Complex Organisms
Of the 35 studies in Table 1, 27 use MTB specific sequences for amplification and/or detection. Six studies use genus specific primers, but the designs preferentially favor amplification of TB complex organisms. Only the study of Irula, I. V. et al., J. Clin. Micro. 31:1811-1814 (1993) focussed on M. avium, and only the study of van der Giessen, J. W. B. et al., J. Clin. Microbiol. 30:1216-1219 (1992) focussed on M. paratuberculosis. Irula used a different isolation technique, making comparisons difficult. However, while PBMC were isolated, they were in fact subjected to a wash step in Tris/EDTA/Triton X-100: if the PBMC had lysed during washing, the bacteria may have been discarded with the supernatant. Regardless, it might be expected that PBMC isolation would be an extremely effective means of capturing the organisms. The study of van der Giessen compared three PCR based systems (McFadden, J. J. et al., Mol. Microbiol. 1:283-291 (1987); van der Giessen, J. W. B. et al., J. Med. Microbiol. 36:255-263 (1992); Vary, P. H. et al., J. Clin. Microbiol. 28:933-937 (1990)) designed to detect M. paratuberculosis in bovine feces (one of these is a commercially available kit from IDEXX (Vary, P. H. et al., J. Clin. Micobiol. 28:933-937 (1990))). Their results were far worse than anything else presented in Table 1 and appear to be artificially low. These results indicate that MAC complex organisms, with their additional lipophilic components, present a further undefined complication to processing for amplification.
J. The Innate Character of the Mycobacteria has Thwarted Exploitation of Amplification Technologies
Clearly, there are two primary sources of false negative results. First, inhibitors are abundant in a variety of specimen types and the preparatory solutions also play a role in modifying the efficiency of the amplification reaction. The second category is due to the innate character of the Mycobacteria. While the source of these characteristics appears obvious, the influence of these idiosyncrasies on sensitivity is so prevalent that Noordhoek et al., J. Clin. Micro. 32:277-284 (1994) conclude: " . . . we will not speculate on the possible factors that might explain the extreme differences in sensitivity of PCR among the seven laboratories . . . . "
The cording, buoyant nature of these organisms causes them to partition in an inefficacious manner, and be poured off with the supernatant. An extreme example of this situation causes "unexplained" results: a sample that is culture positive and smear positive, but appears to be a true negative in the face of multiple amplifications, and does not contain inhibitors. Clearly, the source and nature of these characteristics remains to be fully explained. However, it is these phenomena, the basis of which resides in the innate qualities of these organisms, that the methods described herein solve.